Skip to content

v3.1.0: Strip answer synthesis, keep cognition — production cleanup#23

Merged
Ashish-dwi99 merged 1 commit intomainfrom
alpha
Apr 4, 2026
Merged

v3.1.0: Strip answer synthesis, keep cognition — production cleanup#23
Ashish-dwi99 merged 1 commit intomainfrom
alpha

Conversation

@Ashish-dwi99
Copy link
Copy Markdown
Collaborator

Remove map-reduce answer synthesis (extract_atomic_facts, reduce_atomic_facts, triple decomposition) from query path. Dhee retrieves context; the agent answers.

Write-path enrichment (echo encoding, categories, knowledge graph, scenes, profiles, evolution, distillation) remains default-on — that's Dhee's USP.

  • answer_orchestration.py: 1010 → 228 lines, kept intent classification + query rewriting
  • orchestration.py: 875 → 492 lines, removed _execute_map_reduce, reducer cache
  • Deleted 46 dead files (9 directories): adapters/, api/, decay/, edge/, hive/, integrations/, mini/, retrieval/, teaching/, and 20 core modules
  • Moved DheePlugin → dhee/plugin.py, reranker → dhee/memory/reranker.py
  • Fixed NVIDIA embedder truncation (nemotron 8192 token limit)
  • Added HippoCamp benchmark runner + raw extractors
  • 832 tests passing, 0 failures
  • Net: 169 files → 123 files, ~63K lines → ~51.5K lines

Remove map-reduce answer synthesis (extract_atomic_facts, reduce_atomic_facts,
triple decomposition) from query path. Dhee retrieves context; the agent answers.

Write-path enrichment (echo encoding, categories, knowledge graph, scenes,
profiles, evolution, distillation) remains default-on — that's Dhee's USP.

- answer_orchestration.py: 1010 → 228 lines, kept intent classification + query rewriting
- orchestration.py: 875 → 492 lines, removed _execute_map_reduce, reducer cache
- Deleted 46 dead files (9 directories): adapters/, api/, decay/, edge/,
  hive/, integrations/, mini/, retrieval/, teaching/, and 20 core modules
- Moved DheePlugin → dhee/plugin.py, reranker → dhee/memory/reranker.py
- Fixed NVIDIA embedder truncation (nemotron 8192 token limit)
- Added HippoCamp benchmark runner + raw extractors
- 832 tests passing, 0 failures
- Net: 169 files → 123 files, ~63K lines → ~51.5K lines

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@Ashish-dwi99 Ashish-dwi99 merged commit 8e1a0b5 into main Apr 4, 2026
0 of 3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant